科研成果详情

为缓解软件测试的Oracle问题,蜕变测试通过验证多个测试用例及其输出之间是否满足蜕变关系来验证测试结果.可见蜕变关系是蜕变测试的核心,而现有的蜕变关系检错能力的衡量标准,即错误检出率（Failure Detecting Rate,FDR）是蜕变关系对不同变体的错误检出概率的平均值,这掩盖了蜕变关系的一些重要特征.因而本文提出蜕变关系敏感度的概念,即用蜕变关系对不同变体的错误检出率所构成的多维信息向量,来全面地反映蜕变关系特征,从而为蜕变测试研究提供更多的可能性.蜕变关系敏感度的一个典型应用是对蜕变关系集合进行聚类分析,并挖掘出构建蜕变关系的需求特征与错误发现能力之间的关联.论文使用经典聚类算法k-means进行了实验.实验结果表明,蜕变关系敏感度能很好地支持聚类分析计算,并能挖掘出有用的知识.蜕变关系敏感度为蜕变测试的进一步研究提供了新的方法和依据. 

Metamorphic testing (MT) is proposed to alleviate oracle problem in software testing, which verifies software under testing (SUT) by checking whether inputs and outputs satisfy metamorphic relation (MR). MR, the constraint constructed from the propriety of SUT, plays a key role in MT and becomes focus in MT researching. Failure detecting ratio (FDR) is a popular measure for failure detecting ability of MR. However, FDR is the mean of failure detecting ratios of a MR applied on different mutants. This averaging covers some important properties of MR. In this paper, sensitivity of MR is defined, which is a multi-dimension information vector consisted of failure detecting ratios on all mutants. Sensitivity of MR reflects more fully characteristics of MR than FDR and allows more possibility in MT researching. One typical application of sensitivity of MR is the cluster analysis on MR set, in order to find some clues of the underlying relevant between the requirement of MR and its failure detecting ability. K-means algorithm, a classical cluster analysis algorithm, based on MR sensitivity is applied in the experiment. The experiment results show that the sensitivity of MR supports cluster analysis strongly and the analysis offers some useful knowledge. Sensitivity of MR will be a good method and essential rationale for MT researching.